Phylogenetic Analysis of Characters with Dependencies under Maximum Likelihood

Area of Science:

• Evolutionary Biology
• Computational Biology
• Phylogenetics

Background:

• Phylogenetic analysis often simplifies character dependencies, potentially affecting evolutionary model accuracy.
• Previous methods for handling inapplicable characters and dependencies, like Goloboff et al. (2021), have theoretical limitations.
• Tarasov (2023) proposed synthesizing rate matrices for dependencies, but an efficient method was needed.

Purpose of the Study:

• To develop a more robust and accurate method for incorporating character dependencies into phylogenetic analyses.
• To address the flaws in previous approaches for calculating transition probabilities between morphotypes.
• To provide a computationally efficient way to generate rate matrices for phylogenetic models.

Main Methods:

• Utilized combinations of character states to define ancestral morphotypes, allowing for explicit representation of dependencies.
• Derived transition rate matrices directly from comparisons between legitimate morphotypes, bypassing stepwise synthesis.
• Implemented a user-defined syntax in the TNT program to specify biological dependencies and convert characters into morphotypes.

Main Results:

• Demonstrated that direct comparison of morphotypes yields rate matrices identical to Tarasov's (2023) for inapplicable characters.
• Showed that the product of transition probabilities is flawed and fails time-continuity requirements.
• Established a method for handling complex biological dependencies beyond simple inapplicability.

Conclusions:

• The direct derivation of rate matrices from morphotype comparisons offers a statistically consistent and efficient approach for phylogenetic analysis.
• This method accurately accounts for character dependencies, including inapplicability and complex biological interactions.
• The TNT program's enhanced syntax facilitates the incorporation of sophisticated evolutionary constraints into phylogenetic reconstructions.